2 Linux Ethernet Bonding Driver HOWTO
4 Latest update: 24 April 2006
6 Initial release : Thomas Davis <tadavis at lbl.gov>
7 Corrections, HA extensions : 2000/10/03-15 :
8 - Willy Tarreau <willy at meta-x.org>
9 - Constantine Gavrilov <const-g at xpert.com>
10 - Chad N. Tindel <ctindel at ieee dot org>
11 - Janice Girouard <girouard at us dot ibm dot com>
12 - Jay Vosburgh <fubar at us dot ibm dot com>
14 Reorganized and updated Feb 2005 by Jay Vosburgh
15 Added Sysfs information: 2006/04/24
16 - Mitch Williams <mitch.a.williams at intel.com>
21 The Linux bonding driver provides a method for aggregating
22 multiple network interfaces into a single logical "bonded" interface.
23 The behavior of the bonded interfaces depends upon the mode; generally
24 speaking, modes provide either hot standby or load balancing services.
25 Additionally, link integrity monitoring may be performed.
27 The bonding driver originally came from Donald Becker's
28 beowulf patches for kernel 2.0. It has changed quite a bit since, and
29 the original tools from extreme-linux and beowulf sites will not work
30 with this version of the driver.
32 For new versions of the driver, updated userspace tools, and
33 who to ask for help, please follow the links at the end of this file.
38 1. Bonding Driver Installation
40 2. Bonding Driver Options
42 3. Configuring Bonding Devices
43 3.1 Configuration with Sysconfig Support
44 3.1.1 Using DHCP with Sysconfig
45 3.1.2 Configuring Multiple Bonds with Sysconfig
46 3.2 Configuration with Initscripts Support
47 3.2.1 Using DHCP with Initscripts
48 3.2.2 Configuring Multiple Bonds with Initscripts
49 3.3 Configuring Bonding Manually with Ifenslave
50 3.3.1 Configuring Multiple Bonds Manually
51 3.4 Configuring Bonding Manually via Sysfs
53 4. Querying Bonding Configuration
54 4.1 Bonding Configuration
55 4.2 Network Configuration
57 5. Switch Configuration
59 6. 802.1q VLAN Support
62 7.1 ARP Monitor Operation
63 7.2 Configuring Multiple ARP Targets
64 7.3 MII Monitor Operation
66 8. Potential Trouble Sources
67 8.1 Adventures in Routing
68 8.2 Ethernet Device Renaming
69 8.3 Painfully Slow Or No Failed Link Detection By Miimon
75 11. Configuring Bonding for High Availability
76 11.1 High Availability in a Single Switch Topology
77 11.2 High Availability in a Multiple Switch Topology
78 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
79 11.2.2 HA Link Monitoring for Multiple Switch Topology
81 12. Configuring Bonding for Maximum Throughput
82 12.1 Maximum Throughput in a Single Switch Topology
83 12.1.1 MT Bonding Mode Selection for Single Switch Topology
84 12.1.2 MT Link Monitoring for Single Switch Topology
85 12.2 Maximum Throughput in a Multiple Switch Topology
86 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
87 12.2.2 MT Link Monitoring for Multiple Switch Topology
89 13. Switch Behavior Issues
90 13.1 Link Establishment and Failover Delays
91 13.2 Duplicated Incoming Packets
93 14. Hardware Specific Considerations
96 15. Frequently Asked Questions
98 16. Resources and Links
101 1. Bonding Driver Installation
102 ==============================
104 Most popular distro kernels ship with the bonding driver
105 already available as a module and the ifenslave user level control
106 program installed and ready for use. If your distro does not, or you
107 have need to compile bonding from source (e.g., configuring and
108 installing a mainline kernel from kernel.org), you'll need to perform
111 1.1 Configure and build the kernel with bonding
112 -----------------------------------------------
114 The current version of the bonding driver is available in the
115 drivers/net/bonding subdirectory of the most recent kernel source
116 (which is available on http://kernel.org). Most users "rolling their
117 own" will want to use the most recent kernel from kernel.org.
119 Configure kernel with "make menuconfig" (or "make xconfig" or
120 "make config"), then select "Bonding driver support" in the "Network
121 device support" section. It is recommended that you configure the
122 driver as module since it is currently the only way to pass parameters
123 to the driver or configure more than one bonding device.
125 Build and install the new kernel and modules, then continue
126 below to install ifenslave.
128 1.2 Install ifenslave Control Utility
129 -------------------------------------
131 The ifenslave user level control program is included in the
132 kernel source tree, in the file Documentation/networking/ifenslave.c.
133 It is generally recommended that you use the ifenslave that
134 corresponds to the kernel that you are using (either from the same
135 source tree or supplied with the distro), however, ifenslave
136 executables from older kernels should function (but features newer
137 than the ifenslave release are not supported). Running an ifenslave
138 that is newer than the kernel is not supported, and may or may not
141 To install ifenslave, do the following:
143 # gcc -Wall -O -I/usr/src/linux/include ifenslave.c -o ifenslave
144 # cp ifenslave /sbin/ifenslave
146 If your kernel source is not in "/usr/src/linux," then replace
147 "/usr/src/linux/include" in the above with the location of your kernel
148 source include directory.
150 You may wish to back up any existing /sbin/ifenslave, or, for
151 testing or informal use, tag the ifenslave to the kernel version
152 (e.g., name the ifenslave executable /sbin/ifenslave-2.6.10).
156 If you omit the "-I" or specify an incorrect directory, you
157 may end up with an ifenslave that is incompatible with the kernel
158 you're trying to build it for. Some distros (e.g., Red Hat from 7.1
159 onwards) do not have /usr/include/linux symbolically linked to the
160 default kernel source include directory.
162 SECOND IMPORTANT NOTE:
163 If you plan to configure bonding using sysfs, you do not need
166 2. Bonding Driver Options
167 =========================
169 Options for the bonding driver are supplied as parameters to
170 the bonding module at load time. They may be given as command line
171 arguments to the insmod or modprobe command, but are usually specified
172 in either the /etc/modules.conf or /etc/modprobe.conf configuration
173 file, or in a distro-specific configuration file (some of which are
174 detailed in the next section).
176 The available bonding driver parameters are listed below. If a
177 parameter is not specified the default value is used. When initially
178 configuring a bond, it is recommended "tail -f /var/log/messages" be
179 run in a separate window to watch for bonding driver error messages.
181 It is critical that either the miimon or arp_interval and
182 arp_ip_target parameters be specified, otherwise serious network
183 degradation will occur during link failures. Very few devices do not
184 support at least miimon, so there is really no reason not to use it.
186 Options with textual values will accept either the text name
187 or, for backwards compatibility, the option value. E.g.,
188 "mode=802.3ad" and "mode=4" set the same mode.
190 The parameters are as follows:
194 Specifies the ARP link monitoring frequency in milliseconds.
196 The ARP monitor works by periodically checking the slave
197 devices to determine whether they have sent or received
198 traffic recently (the precise criteria depends upon the
199 bonding mode, and the state of the slave). Regular traffic is
200 generated via ARP probes issued for the addresses specified by
201 the arp_ip_target option.
203 This behavior can be modified by the arp_validate option,
206 If ARP monitoring is used in an etherchannel compatible mode
207 (modes 0 and 2), the switch should be configured in a mode
208 that evenly distributes packets across all links. If the
209 switch is configured to distribute the packets in an XOR
210 fashion, all replies from the ARP targets will be received on
211 the same link which could cause the other team members to
212 fail. ARP monitoring should not be used in conjunction with
213 miimon. A value of 0 disables ARP monitoring. The default
218 Specifies the IP addresses to use as ARP monitoring peers when
219 arp_interval is > 0. These are the targets of the ARP request
220 sent to determine the health of the link to the targets.
221 Specify these values in ddd.ddd.ddd.ddd format. Multiple IP
222 addresses must be separated by a comma. At least one IP
223 address must be given for ARP monitoring to function. The
224 maximum number of targets that can be specified is 16. The
225 default value is no IP addresses.
229 Specifies whether or not ARP probes and replies should be
230 validated in the active-backup mode. This causes the ARP
231 monitor to examine the incoming ARP requests and replies, and
232 only consider a slave to be up if it is receiving the
233 appropriate ARP traffic.
239 No validation is performed. This is the default.
243 Validation is performed only for the active slave.
247 Validation is performed only for backup slaves.
251 Validation is performed for all slaves.
253 For the active slave, the validation checks ARP replies to
254 confirm that they were generated by an arp_ip_target. Since
255 backup slaves do not typically receive these replies, the
256 validation performed for backup slaves is on the ARP request
257 sent out via the active slave. It is possible that some
258 switch or network configurations may result in situations
259 wherein the backup slaves do not receive the ARP requests; in
260 such a situation, validation of backup slaves must be
263 This option is useful in network configurations in which
264 multiple bonding hosts are concurrently issuing ARPs to one or
265 more targets beyond a common switch. Should the link between
266 the switch and target fail (but not the switch itself), the
267 probe traffic generated by the multiple bonding instances will
268 fool the standard ARP monitor into considering the links as
269 still up. Use of the arp_validate option can resolve this, as
270 the ARP monitor will only consider ARP requests and replies
271 associated with its own instance of bonding.
273 This option was added in bonding version 3.1.0.
277 Specifies the time, in milliseconds, to wait before disabling
278 a slave after a link failure has been detected. This option
279 is only valid for the miimon link monitor. The downdelay
280 value should be a multiple of the miimon value; if not, it
281 will be rounded down to the nearest multiple. The default
286 Specifies whether active-backup mode should set all slaves to
287 the same MAC address (the traditional behavior), or, when
288 enabled, change the bond's MAC address when changing the
289 active interface (i.e., fail over the MAC address itself).
291 Fail over MAC is useful for devices that cannot ever alter
292 their MAC address, or for devices that refuse incoming
293 broadcasts with their own source MAC (which interferes with
296 The down side of fail over MAC is that every device on the
297 network must be updated via gratuitous ARP, vs. just updating
298 a switch or set of switches (which often takes place for any
299 traffic, not just ARP traffic, if the switch snoops incoming
300 traffic to update its tables) for the traditional method. If
301 the gratuitous ARP is lost, communication may be disrupted.
303 When fail over MAC is used in conjuction with the mii monitor,
304 devices which assert link up prior to being able to actually
305 transmit and receive are particularly susecptible to loss of
306 the gratuitous ARP, and an appropriate updelay setting may be
309 A value of 0 disables fail over MAC, and is the default. A
310 value of 1 enables fail over MAC. This option is enabled
311 automatically if the first slave added cannot change its MAC
312 address. This option may be modified via sysfs only when no
313 slaves are present in the bond.
315 This option was added in bonding version 3.2.0.
319 Option specifying the rate in which we'll ask our link partner
320 to transmit LACPDU packets in 802.3ad mode. Possible values
324 Request partner to transmit LACPDUs every 30 seconds
327 Request partner to transmit LACPDUs every 1 second
333 Specifies the number of bonding devices to create for this
334 instance of the bonding driver. E.g., if max_bonds is 3, and
335 the bonding driver is not already loaded, then bond0, bond1
336 and bond2 will be created. The default value is 1.
340 Specifies the MII link monitoring frequency in milliseconds.
341 This determines how often the link state of each slave is
342 inspected for link failures. A value of zero disables MII
343 link monitoring. A value of 100 is a good starting point.
344 The use_carrier option, below, affects how the link state is
345 determined. See the High Availability section for additional
346 information. The default value is 0.
350 Specifies one of the bonding policies. The default is
351 balance-rr (round robin). Possible values are:
355 Round-robin policy: Transmit packets in sequential
356 order from the first available slave through the
357 last. This mode provides load balancing and fault
362 Active-backup policy: Only one slave in the bond is
363 active. A different slave becomes active if, and only
364 if, the active slave fails. The bond's MAC address is
365 externally visible on only one port (network adapter)
366 to avoid confusing the switch.
368 In bonding version 2.6.2 or later, when a failover
369 occurs in active-backup mode, bonding will issue one
370 or more gratuitous ARPs on the newly active slave.
371 One gratuitous ARP is issued for the bonding master
372 interface and each VLAN interfaces configured above
373 it, provided that the interface has at least one IP
374 address configured. Gratuitous ARPs issued for VLAN
375 interfaces are tagged with the appropriate VLAN id.
377 This mode provides fault tolerance. The primary
378 option, documented below, affects the behavior of this
383 XOR policy: Transmit based on the selected transmit
384 hash policy. The default policy is a simple [(source
385 MAC address XOR'd with destination MAC address) modulo
386 slave count]. Alternate transmit policies may be
387 selected via the xmit_hash_policy option, described
390 This mode provides load balancing and fault tolerance.
394 Broadcast policy: transmits everything on all slave
395 interfaces. This mode provides fault tolerance.
399 IEEE 802.3ad Dynamic link aggregation. Creates
400 aggregation groups that share the same speed and
401 duplex settings. Utilizes all slaves in the active
402 aggregator according to the 802.3ad specification.
404 Slave selection for outgoing traffic is done according
405 to the transmit hash policy, which may be changed from
406 the default simple XOR policy via the xmit_hash_policy
407 option, documented below. Note that not all transmit
408 policies may be 802.3ad compliant, particularly in
409 regards to the packet mis-ordering requirements of
410 section 43.2.4 of the 802.3ad standard. Differing
411 peer implementations will have varying tolerances for
416 1. Ethtool support in the base drivers for retrieving
417 the speed and duplex of each slave.
419 2. A switch that supports IEEE 802.3ad Dynamic link
422 Most switches will require some type of configuration
423 to enable 802.3ad mode.
427 Adaptive transmit load balancing: channel bonding that
428 does not require any special switch support. The
429 outgoing traffic is distributed according to the
430 current load (computed relative to the speed) on each
431 slave. Incoming traffic is received by the current
432 slave. If the receiving slave fails, another slave
433 takes over the MAC address of the failed receiving
438 Ethtool support in the base drivers for retrieving the
443 Adaptive load balancing: includes balance-tlb plus
444 receive load balancing (rlb) for IPV4 traffic, and
445 does not require any special switch support. The
446 receive load balancing is achieved by ARP negotiation.
447 The bonding driver intercepts the ARP Replies sent by
448 the local system on their way out and overwrites the
449 source hardware address with the unique hardware
450 address of one of the slaves in the bond such that
451 different peers use different hardware addresses for
454 Receive traffic from connections created by the server
455 is also balanced. When the local system sends an ARP
456 Request the bonding driver copies and saves the peer's
457 IP information from the ARP packet. When the ARP
458 Reply arrives from the peer, its hardware address is
459 retrieved and the bonding driver initiates an ARP
460 reply to this peer assigning it to one of the slaves
461 in the bond. A problematic outcome of using ARP
462 negotiation for balancing is that each time that an
463 ARP request is broadcast it uses the hardware address
464 of the bond. Hence, peers learn the hardware address
465 of the bond and the balancing of receive traffic
466 collapses to the current slave. This is handled by
467 sending updates (ARP Replies) to all the peers with
468 their individually assigned hardware address such that
469 the traffic is redistributed. Receive traffic is also
470 redistributed when a new slave is added to the bond
471 and when an inactive slave is re-activated. The
472 receive load is distributed sequentially (round robin)
473 among the group of highest speed slaves in the bond.
475 When a link is reconnected or a new slave joins the
476 bond the receive traffic is redistributed among all
477 active slaves in the bond by initiating ARP Replies
478 with the selected MAC address to each of the
479 clients. The updelay parameter (detailed below) must
480 be set to a value equal or greater than the switch's
481 forwarding delay so that the ARP Replies sent to the
482 peers will not be blocked by the switch.
486 1. Ethtool support in the base drivers for retrieving
487 the speed of each slave.
489 2. Base driver support for setting the hardware
490 address of a device while it is open. This is
491 required so that there will always be one slave in the
492 team using the bond hardware address (the
493 curr_active_slave) while having a unique hardware
494 address for each slave in the bond. If the
495 curr_active_slave fails its hardware address is
496 swapped with the new curr_active_slave that was
501 A string (eth0, eth2, etc) specifying which slave is the
502 primary device. The specified device will always be the
503 active slave while it is available. Only when the primary is
504 off-line will alternate devices be used. This is useful when
505 one slave is preferred over another, e.g., when one slave has
506 higher throughput than another.
508 The primary option is only valid for active-backup mode.
512 Specifies the time, in milliseconds, to wait before enabling a
513 slave after a link recovery has been detected. This option is
514 only valid for the miimon link monitor. The updelay value
515 should be a multiple of the miimon value; if not, it will be
516 rounded down to the nearest multiple. The default value is 0.
520 Specifies whether or not miimon should use MII or ETHTOOL
521 ioctls vs. netif_carrier_ok() to determine the link
522 status. The MII or ETHTOOL ioctls are less efficient and
523 utilize a deprecated calling sequence within the kernel. The
524 netif_carrier_ok() relies on the device driver to maintain its
525 state with netif_carrier_on/off; at this writing, most, but
526 not all, device drivers support this facility.
528 If bonding insists that the link is up when it should not be,
529 it may be that your network device driver does not support
530 netif_carrier_on/off. The default state for netif_carrier is
531 "carrier on," so if a driver does not support netif_carrier,
532 it will appear as if the link is always up. In this case,
533 setting use_carrier to 0 will cause bonding to revert to the
534 MII / ETHTOOL ioctl method to determine the link state.
536 A value of 1 enables the use of netif_carrier_ok(), a value of
537 0 will use the deprecated MII / ETHTOOL ioctls. The default
542 Selects the transmit hash policy to use for slave selection in
543 balance-xor and 802.3ad modes. Possible values are:
547 Uses XOR of hardware MAC addresses to generate the
550 (source MAC XOR destination MAC) modulo slave count
552 This algorithm will place all traffic to a particular
553 network peer on the same slave.
555 This algorithm is 802.3ad compliant.
559 This policy uses a combination of layer2 and layer3
560 protocol information to generate the hash.
562 Uses XOR of hardware MAC addresses and IP addresses to
563 generate the hash. The formula is
565 (((source IP XOR dest IP) AND 0xffff) XOR
566 ( source MAC XOR destination MAC ))
569 This algorithm will place all traffic to a particular
570 network peer on the same slave. For non-IP traffic,
571 the formula is the same as for the layer2 transmit
574 This policy is intended to provide a more balanced
575 distribution of traffic than layer2 alone, especially
576 in environments where a layer3 gateway device is
577 required to reach most destinations.
579 This algorithm is 802.3ad complient.
583 This policy uses upper layer protocol information,
584 when available, to generate the hash. This allows for
585 traffic to a particular network peer to span multiple
586 slaves, although a single connection will not span
589 The formula for unfragmented TCP and UDP packets is
591 ((source port XOR dest port) XOR
592 ((source IP XOR dest IP) AND 0xffff)
595 For fragmented TCP or UDP packets and all other IP
596 protocol traffic, the source and destination port
597 information is omitted. For non-IP traffic, the
598 formula is the same as for the layer2 transmit hash
601 This policy is intended to mimic the behavior of
602 certain switches, notably Cisco switches with PFC2 as
603 well as some Foundry and IBM products.
605 This algorithm is not fully 802.3ad compliant. A
606 single TCP or UDP conversation containing both
607 fragmented and unfragmented packets will see packets
608 striped across two interfaces. This may result in out
609 of order delivery. Most traffic types will not meet
610 this criteria, as TCP rarely fragments traffic, and
611 most UDP traffic is not involved in extended
612 conversations. Other implementations of 802.3ad may
613 or may not tolerate this noncompliance.
615 The default value is layer2. This option was added in bonding
616 version 2.6.3. In earlier versions of bonding, this parameter
617 does not exist, and the layer2 policy is the only policy. The
618 layer2+3 value was added for bonding version 3.2.2.
621 3. Configuring Bonding Devices
622 ==============================
624 You can configure bonding using either your distro's network
625 initialization scripts, or manually using either ifenslave or the
626 sysfs interface. Distros generally use one of two packages for the
627 network initialization scripts: initscripts or sysconfig. Recent
628 versions of these packages have support for bonding, while older
631 We will first describe the options for configuring bonding for
632 distros using versions of initscripts and sysconfig with full or
633 partial support for bonding, then provide information on enabling
634 bonding without support from the network initialization scripts (i.e.,
635 older versions of initscripts or sysconfig).
637 If you're unsure whether your distro uses sysconfig or
638 initscripts, or don't know if it's new enough, have no fear.
639 Determining this is fairly straightforward.
641 First, issue the command:
645 It will respond with a line of text starting with either
646 "initscripts" or "sysconfig," followed by some numbers. This is the
647 package that provides your network initialization scripts.
649 Next, to determine if your installation supports bonding,
652 $ grep ifenslave /sbin/ifup
654 If this returns any matches, then your initscripts or
655 sysconfig has support for bonding.
657 3.1 Configuration with Sysconfig Support
658 ----------------------------------------
660 This section applies to distros using a version of sysconfig
661 with bonding support, for example, SuSE Linux Enterprise Server 9.
663 SuSE SLES 9's networking configuration system does support
664 bonding, however, at this writing, the YaST system configuration
665 front end does not provide any means to work with bonding devices.
666 Bonding devices can be managed by hand, however, as follows.
668 First, if they have not already been configured, configure the
669 slave devices. On SLES 9, this is most easily done by running the
670 yast2 sysconfig configuration utility. The goal is for to create an
671 ifcfg-id file for each slave device. The simplest way to accomplish
672 this is to configure the devices for DHCP (this is only to get the
673 file ifcfg-id file created; see below for some issues with DHCP). The
674 name of the configuration file for each device will be of the form:
676 ifcfg-id-xx:xx:xx:xx:xx:xx
678 Where the "xx" portion will be replaced with the digits from
679 the device's permanent MAC address.
681 Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
682 created, it is necessary to edit the configuration files for the slave
683 devices (the MAC addresses correspond to those of the slave devices).
684 Before editing, the file will contain multiple lines, and will look
690 UNIQUE='XNzu.WeZGOGF+4wE'
691 _nm_name='bus-pci-0001:61:01.0'
693 Change the BOOTPROTO and STARTMODE lines to the following:
698 Do not alter the UNIQUE or _nm_name lines. Remove any other
699 lines (USERCTL, etc).
701 Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
702 it's time to create the configuration file for the bonding device
703 itself. This file is named ifcfg-bondX, where X is the number of the
704 bonding device to create, starting at 0. The first such file is
705 ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig
706 network configuration system will correctly start multiple instances
709 The contents of the ifcfg-bondX file is as follows:
712 BROADCAST="10.0.2.255"
714 NETMASK="255.255.0.0"
719 BONDING_MODULE_OPTS="mode=active-backup miimon=100"
720 BONDING_SLAVE0="eth0"
721 BONDING_SLAVE1="bus-pci-0000:06:08.1"
723 Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
724 values with the appropriate values for your network.
726 The STARTMODE specifies when the device is brought online.
727 The possible values are:
729 onboot: The device is started at boot time. If you're not
730 sure, this is probably what you want.
732 manual: The device is started only when ifup is called
733 manually. Bonding devices may be configured this
734 way if you do not wish them to start automatically
735 at boot for some reason.
737 hotplug: The device is started by a hotplug event. This is not
738 a valid choice for a bonding device.
740 off or ignore: The device configuration is ignored.
742 The line BONDING_MASTER='yes' indicates that the device is a
743 bonding master device. The only useful value is "yes."
745 The contents of BONDING_MODULE_OPTS are supplied to the
746 instance of the bonding module for this device. Specify the options
747 for the bonding mode, link monitoring, and so on here. Do not include
748 the max_bonds bonding parameter; this will confuse the configuration
749 system if you have multiple bonding devices.
751 Finally, supply one BONDING_SLAVEn="slave device" for each
752 slave. where "n" is an increasing value, one for each slave. The
753 "slave device" is either an interface name, e.g., "eth0", or a device
754 specifier for the network device. The interface name is easier to
755 find, but the ethN names are subject to change at boot time if, e.g.,
756 a device early in the sequence has failed. The device specifiers
757 (bus-pci-0000:06:08.1 in the example above) specify the physical
758 network device, and will not change unless the device's bus location
759 changes (for example, it is moved from one PCI slot to another). The
760 example above uses one of each type for demonstration purposes; most
761 configurations will choose one or the other for all slave devices.
763 When all configuration files have been modified or created,
764 networking must be restarted for the configuration changes to take
765 effect. This can be accomplished via the following:
767 # /etc/init.d/network restart
769 Note that the network control script (/sbin/ifdown) will
770 remove the bonding module as part of the network shutdown processing,
771 so it is not necessary to remove the module by hand if, e.g., the
772 module parameters have changed.
774 Also, at this writing, YaST/YaST2 will not manage bonding
775 devices (they do not show bonding interfaces on its list of network
776 devices). It is necessary to edit the configuration file by hand to
777 change the bonding configuration.
779 Additional general options and details of the ifcfg file
780 format can be found in an example ifcfg template file:
782 /etc/sysconfig/network/ifcfg.template
784 Note that the template does not document the various BONDING_
785 settings described above, but does describe many of the other options.
787 3.1.1 Using DHCP with Sysconfig
788 -------------------------------
790 Under sysconfig, configuring a device with BOOTPROTO='dhcp'
791 will cause it to query DHCP for its IP address information. At this
792 writing, this does not function for bonding devices; the scripts
793 attempt to obtain the device address from DHCP prior to adding any of
794 the slave devices. Without active slaves, the DHCP requests are not
797 3.1.2 Configuring Multiple Bonds with Sysconfig
798 -----------------------------------------------
800 The sysconfig network initialization system is capable of
801 handling multiple bonding devices. All that is necessary is for each
802 bonding instance to have an appropriately configured ifcfg-bondX file
803 (as described above). Do not specify the "max_bonds" parameter to any
804 instance of bonding, as this will confuse sysconfig. If you require
805 multiple bonding devices with identical parameters, create multiple
808 Because the sysconfig scripts supply the bonding module
809 options in the ifcfg-bondX file, it is not necessary to add them to
810 the system /etc/modules.conf or /etc/modprobe.conf configuration file.
812 3.2 Configuration with Initscripts Support
813 ------------------------------------------
815 This section applies to distros using a version of initscripts
816 with bonding support, for example, Red Hat Linux 9 or Red Hat
817 Enterprise Linux version 3 or 4. On these systems, the network
818 initialization scripts have some knowledge of bonding, and can be
819 configured to control bonding devices.
821 These distros will not automatically load the network adapter
822 driver unless the ethX device is configured with an IP address.
823 Because of this constraint, users must manually configure a
824 network-script file for all physical adapters that will be members of
825 a bondX link. Network script files are located in the directory:
827 /etc/sysconfig/network-scripts
829 The file name must be prefixed with "ifcfg-eth" and suffixed
830 with the adapter's physical adapter number. For example, the script
831 for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
832 Place the following text in the file:
841 The DEVICE= line will be different for every ethX device and
842 must correspond with the name of the file, i.e., ifcfg-eth1 must have
843 a device line of DEVICE=eth1. The setting of the MASTER= line will
844 also depend on the final bonding interface name chosen for your bond.
845 As with other network devices, these typically start at 0, and go up
846 one for each device, i.e., the first bonding instance is bond0, the
847 second is bond1, and so on.
849 Next, create a bond network script. The file name for this
850 script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
851 the number of the bond. For bond0 the file is named "ifcfg-bond0",
852 for bond1 it is named "ifcfg-bond1", and so on. Within that file,
853 place the following text:
857 NETMASK=255.255.255.0
859 BROADCAST=192.168.1.255
864 Be sure to change the networking specific lines (IPADDR,
865 NETMASK, NETWORK and BROADCAST) to match your network configuration.
867 Finally, it is necessary to edit /etc/modules.conf (or
868 /etc/modprobe.conf, depending upon your distro) to load the bonding
869 module with your desired options when the bond0 interface is brought
870 up. The following lines in /etc/modules.conf (or modprobe.conf) will
871 load the bonding module, and select its options:
874 options bond0 mode=balance-alb miimon=100
876 Replace the sample parameters with the appropriate set of
877 options for your configuration.
879 Finally run "/etc/rc.d/init.d/network restart" as root. This
880 will restart the networking subsystem and your bond link should be now
883 3.2.1 Using DHCP with Initscripts
884 ---------------------------------
886 Recent versions of initscripts (the version supplied with
887 Fedora Core 3 and Red Hat Enterprise Linux 4 is reported to work) do
888 have support for assigning IP information to bonding devices via DHCP.
890 To configure bonding for DHCP, configure it as described
891 above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
892 and add a line consisting of "TYPE=Bonding". Note that the TYPE value
895 3.2.2 Configuring Multiple Bonds with Initscripts
896 -------------------------------------------------
898 At this writing, the initscripts package does not directly
899 support loading the bonding driver multiple times, so the process for
900 doing so is the same as described in the "Configuring Multiple Bonds
901 Manually" section, below.
903 NOTE: It has been observed that some Red Hat supplied kernels
904 are apparently unable to rename modules at load time (the "-o bond1"
905 part). Attempts to pass that option to modprobe will produce an
906 "Operation not permitted" error. This has been reported on some
907 Fedora Core kernels, and has been seen on RHEL 4 as well. On kernels
908 exhibiting this problem, it will be impossible to configure multiple
909 bonds with differing parameters.
911 3.3 Configuring Bonding Manually with Ifenslave
912 -----------------------------------------------
914 This section applies to distros whose network initialization
915 scripts (the sysconfig or initscripts package) do not have specific
916 knowledge of bonding. One such distro is SuSE Linux Enterprise Server
919 The general method for these systems is to place the bonding
920 module parameters into /etc/modules.conf or /etc/modprobe.conf (as
921 appropriate for the installed distro), then add modprobe and/or
922 ifenslave commands to the system's global init script. The name of
923 the global init script differs; for sysconfig, it is
924 /etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
926 For example, if you wanted to make a simple bond of two e100
927 devices (presumed to be eth0 and eth1), and have it persist across
928 reboots, edit the appropriate file (/etc/init.d/boot.local or
929 /etc/rc.d/rc.local), and add the following:
931 modprobe bonding mode=balance-alb miimon=100
933 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
937 Replace the example bonding module parameters and bond0
938 network configuration (IP address, netmask, etc) with the appropriate
939 values for your configuration.
941 Unfortunately, this method will not provide support for the
942 ifup and ifdown scripts on the bond devices. To reload the bonding
943 configuration, it is necessary to run the initialization script, e.g.,
945 # /etc/init.d/boot.local
951 It may be desirable in such a case to create a separate script
952 which only initializes the bonding configuration, then call that
953 separate script from within boot.local. This allows for bonding to be
954 enabled without re-running the entire global init script.
956 To shut down the bonding devices, it is necessary to first
957 mark the bonding device itself as being down, then remove the
958 appropriate device driver modules. For our example above, you can do
961 # ifconfig bond0 down
965 Again, for convenience, it may be desirable to create a script
969 3.3.1 Configuring Multiple Bonds Manually
970 -----------------------------------------
972 This section contains information on configuring multiple
973 bonding devices with differing options for those systems whose network
974 initialization scripts lack support for configuring multiple bonds.
976 If you require multiple bonding devices, but all with the same
977 options, you may wish to use the "max_bonds" module parameter,
980 To create multiple bonding devices with differing options, it
981 is necessary to use bonding parameters exported by sysfs, documented
982 in the section below.
985 3.4 Configuring Bonding Manually via Sysfs
986 ------------------------------------------
988 Starting with version 3.0, Channel Bonding may be configured
989 via the sysfs interface. This interface allows dynamic configuration
990 of all bonds in the system without unloading the module. It also
991 allows for adding and removing bonds at runtime. Ifenslave is no
992 longer required, though it is still supported.
994 Use of the sysfs interface allows you to use multiple bonds
995 with different configurations without having to reload the module.
996 It also allows you to use multiple, differently configured bonds when
997 bonding is compiled into the kernel.
999 You must have the sysfs filesystem mounted to configure
1000 bonding this way. The examples in this document assume that you
1001 are using the standard mount point for sysfs, e.g. /sys. If your
1002 sysfs filesystem is mounted elsewhere, you will need to adjust the
1003 example paths accordingly.
1005 Creating and Destroying Bonds
1006 -----------------------------
1007 To add a new bond foo:
1008 # echo +foo > /sys/class/net/bonding_masters
1010 To remove an existing bond bar:
1011 # echo -bar > /sys/class/net/bonding_masters
1013 To show all existing bonds:
1014 # cat /sys/class/net/bonding_masters
1016 NOTE: due to 4K size limitation of sysfs files, this list may be
1017 truncated if you have more than a few hundred bonds. This is unlikely
1018 to occur under normal operating conditions.
1020 Adding and Removing Slaves
1021 --------------------------
1022 Interfaces may be enslaved to a bond using the file
1023 /sys/class/net/<bond>/bonding/slaves. The semantics for this file
1024 are the same as for the bonding_masters file.
1026 To enslave interface eth0 to bond bond0:
1028 # echo +eth0 > /sys/class/net/bond0/bonding/slaves
1030 To free slave eth0 from bond bond0:
1031 # echo -eth0 > /sys/class/net/bond0/bonding/slaves
1033 NOTE: The bond must be up before slaves can be added. All
1034 slaves are freed when the interface is brought down.
1036 When an interface is enslaved to a bond, symlinks between the
1037 two are created in the sysfs filesystem. In this case, you would get
1038 /sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1039 /sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1041 This means that you can tell quickly whether or not an
1042 interface is enslaved by looking for the master symlink. Thus:
1043 # echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1044 will free eth0 from whatever bond it is enslaved to, regardless of
1045 the name of the bond interface.
1047 Changing a Bond's Configuration
1048 -------------------------------
1049 Each bond may be configured individually by manipulating the
1050 files located in /sys/class/net/<bond name>/bonding
1052 The names of these files correspond directly with the command-
1053 line parameters described elsewhere in this file, and, with the
1054 exception of arp_ip_target, they accept the same values. To see the
1055 current setting, simply cat the appropriate file.
1057 A few examples will be given here; for specific usage
1058 guidelines for each parameter, see the appropriate section in this
1061 To configure bond0 for balance-alb mode:
1062 # ifconfig bond0 down
1063 # echo 6 > /sys/class/net/bond0/bonding/mode
1065 # echo balance-alb > /sys/class/net/bond0/bonding/mode
1066 NOTE: The bond interface must be down before the mode can be
1069 To enable MII monitoring on bond0 with a 1 second interval:
1070 # echo 1000 > /sys/class/net/bond0/bonding/miimon
1071 NOTE: If ARP monitoring is enabled, it will disabled when MII
1072 monitoring is enabled, and vice-versa.
1075 # echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1076 # echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
1077 NOTE: up to 10 target addresses may be specified.
1079 To remove an ARP target:
1080 # echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1082 Example Configuration
1083 ---------------------
1084 We begin with the same example that is shown in section 3.3,
1085 executed with sysfs, and without using ifenslave.
1087 To make a simple bond of two e100 devices (presumed to be eth0
1088 and eth1), and have it persist across reboots, edit the appropriate
1089 file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1094 echo balance-alb > /sys/class/net/bond0/bonding/mode
1095 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1096 echo 100 > /sys/class/net/bond0/bonding/miimon
1097 echo +eth0 > /sys/class/net/bond0/bonding/slaves
1098 echo +eth1 > /sys/class/net/bond0/bonding/slaves
1100 To add a second bond, with two e1000 interfaces in
1101 active-backup mode, using ARP monitoring, add the following lines to
1105 echo +bond1 > /sys/class/net/bonding_masters
1106 echo active-backup > /sys/class/net/bond1/bonding/mode
1107 ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1108 echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1109 echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1110 echo +eth2 > /sys/class/net/bond1/bonding/slaves
1111 echo +eth3 > /sys/class/net/bond1/bonding/slaves
1114 4. Querying Bonding Configuration
1115 =================================
1117 4.1 Bonding Configuration
1118 -------------------------
1120 Each bonding device has a read-only file residing in the
1121 /proc/net/bonding directory. The file contents include information
1122 about the bonding configuration, options and state of each slave.
1124 For example, the contents of /proc/net/bonding/bond0 after the
1125 driver is loaded with parameters of mode=0 and miimon=1000 is
1126 generally as follows:
1128 Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1129 Bonding Mode: load balancing (round-robin)
1130 Currently Active Slave: eth0
1132 MII Polling Interval (ms): 1000
1136 Slave Interface: eth1
1138 Link Failure Count: 1
1140 Slave Interface: eth0
1142 Link Failure Count: 1
1144 The precise format and contents will change depending upon the
1145 bonding configuration, state, and version of the bonding driver.
1147 4.2 Network configuration
1148 -------------------------
1150 The network configuration can be inspected using the ifconfig
1151 command. Bonding devices will have the MASTER flag set; Bonding slave
1152 devices will have the SLAVE flag set. The ifconfig output does not
1153 contain information on which slaves are associated with which masters.
1155 In the example below, the bond0 interface is the master
1156 (MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1157 bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1158 TLB and ALB that require a unique MAC address for each slave.
1161 bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1162 inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0
1163 UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1
1164 RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1165 TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1166 collisions:0 txqueuelen:0
1168 eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1169 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1170 RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1171 TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1172 collisions:0 txqueuelen:100
1173 Interrupt:10 Base address:0x1080
1175 eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1176 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1177 RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1178 TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1179 collisions:0 txqueuelen:100
1180 Interrupt:9 Base address:0x1400
1182 5. Switch Configuration
1183 =======================
1185 For this section, "switch" refers to whatever system the
1186 bonded devices are directly connected to (i.e., where the other end of
1187 the cable plugs into). This may be an actual dedicated switch device,
1188 or it may be another regular system (e.g., another computer running
1191 The active-backup, balance-tlb and balance-alb modes do not
1192 require any specific configuration of the switch.
1194 The 802.3ad mode requires that the switch have the appropriate
1195 ports configured as an 802.3ad aggregation. The precise method used
1196 to configure this varies from switch to switch, but, for example, a
1197 Cisco 3550 series switch requires that the appropriate ports first be
1198 grouped together in a single etherchannel instance, then that
1199 etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1200 standard EtherChannel).
1202 The balance-rr, balance-xor and broadcast modes generally
1203 require that the switch have the appropriate ports grouped together.
1204 The nomenclature for such a group differs between switches, it may be
1205 called an "etherchannel" (as in the Cisco example, above), a "trunk
1206 group" or some other similar variation. For these modes, each switch
1207 will also have its own configuration options for the switch's transmit
1208 policy to the bond. Typical choices include XOR of either the MAC or
1209 IP addresses. The transmit policy of the two peers does not need to
1210 match. For these three modes, the bonding mode really selects a
1211 transmit policy for an EtherChannel group; all three will interoperate
1212 with another EtherChannel group.
1215 6. 802.1q VLAN Support
1216 ======================
1218 It is possible to configure VLAN devices over a bond interface
1219 using the 8021q driver. However, only packets coming from the 8021q
1220 driver and passing through bonding will be tagged by default. Self
1221 generated packets, for example, bonding's learning packets or ARP
1222 packets generated by either ALB mode or the ARP monitor mechanism, are
1223 tagged internally by bonding itself. As a result, bonding must
1224 "learn" the VLAN IDs configured above it, and use those IDs to tag
1225 self generated packets.
1227 For reasons of simplicity, and to support the use of adapters
1228 that can do VLAN hardware acceleration offloading, the bonding
1229 interface declares itself as fully hardware offloading capable, it gets
1230 the add_vid/kill_vid notifications to gather the necessary
1231 information, and it propagates those actions to the slaves. In case
1232 of mixed adapter types, hardware accelerated tagged packets that
1233 should go through an adapter that is not offloading capable are
1234 "un-accelerated" by the bonding driver so the VLAN tag sits in the
1237 VLAN interfaces *must* be added on top of a bonding interface
1238 only after enslaving at least one slave. The bonding interface has a
1239 hardware address of 00:00:00:00:00:00 until the first slave is added.
1240 If the VLAN interface is created prior to the first enslavement, it
1241 would pick up the all-zeroes hardware address. Once the first slave
1242 is attached to the bond, the bond device itself will pick up the
1243 slave's hardware address, which is then available for the VLAN device.
1245 Also, be aware that a similar problem can occur if all slaves
1246 are released from a bond that still has one or more VLAN interfaces on
1247 top of it. When a new slave is added, the bonding interface will
1248 obtain its hardware address from the first slave, which might not
1249 match the hardware address of the VLAN interfaces (which was
1250 ultimately copied from an earlier slave).
1252 There are two methods to insure that the VLAN device operates
1253 with the correct hardware address if all slaves are removed from a
1256 1. Remove all VLAN interfaces then recreate them
1258 2. Set the bonding interface's hardware address so that it
1259 matches the hardware address of the VLAN interfaces.
1261 Note that changing a VLAN interface's HW address would set the
1262 underlying device -- i.e. the bonding interface -- to promiscuous
1263 mode, which might not be what you want.
1269 The bonding driver at present supports two schemes for
1270 monitoring a slave device's link state: the ARP monitor and the MII
1273 At the present time, due to implementation restrictions in the
1274 bonding driver itself, it is not possible to enable both ARP and MII
1275 monitoring simultaneously.
1277 7.1 ARP Monitor Operation
1278 -------------------------
1280 The ARP monitor operates as its name suggests: it sends ARP
1281 queries to one or more designated peer systems on the network, and
1282 uses the response as an indication that the link is operating. This
1283 gives some assurance that traffic is actually flowing to and from one
1284 or more peers on the local network.
1286 The ARP monitor relies on the device driver itself to verify
1287 that traffic is flowing. In particular, the driver must keep up to
1288 date the last receive time, dev->last_rx, and transmit start time,
1289 dev->trans_start. If these are not updated by the driver, then the
1290 ARP monitor will immediately fail any slaves using that driver, and
1291 those slaves will stay down. If networking monitoring (tcpdump, etc)
1292 shows the ARP requests and replies on the network, then it may be that
1293 your device driver is not updating last_rx and trans_start.
1295 7.2 Configuring Multiple ARP Targets
1296 ------------------------------------
1298 While ARP monitoring can be done with just one target, it can
1299 be useful in a High Availability setup to have several targets to
1300 monitor. In the case of just one target, the target itself may go
1301 down or have a problem making it unresponsive to ARP requests. Having
1302 an additional target (or several) increases the reliability of the ARP
1305 Multiple ARP targets must be separated by commas as follows:
1307 # example options for ARP monitoring with three targets
1309 options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1311 For just a single target the options would resemble:
1313 # example options for ARP monitoring with one target
1315 options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1318 7.3 MII Monitor Operation
1319 -------------------------
1321 The MII monitor monitors only the carrier state of the local
1322 network interface. It accomplishes this in one of three ways: by
1323 depending upon the device driver to maintain its carrier state, by
1324 querying the device's MII registers, or by making an ethtool query to
1327 If the use_carrier module parameter is 1 (the default value),
1328 then the MII monitor will rely on the driver for carrier state
1329 information (via the netif_carrier subsystem). As explained in the
1330 use_carrier parameter information, above, if the MII monitor fails to
1331 detect carrier loss on the device (e.g., when the cable is physically
1332 disconnected), it may be that the driver does not support
1335 If use_carrier is 0, then the MII monitor will first query the
1336 device's (via ioctl) MII registers and check the link state. If that
1337 request fails (not just that it returns carrier down), then the MII
1338 monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain
1339 the same information. If both methods fail (i.e., the driver either
1340 does not support or had some error in processing both the MII register
1341 and ethtool requests), then the MII monitor will assume the link is
1344 8. Potential Sources of Trouble
1345 ===============================
1347 8.1 Adventures in Routing
1348 -------------------------
1350 When bonding is configured, it is important that the slave
1351 devices not have routes that supersede routes of the master (or,
1352 generally, not have routes at all). For example, suppose the bonding
1353 device bond0 has two slaves, eth0 and eth1, and the routing table is
1356 Kernel IP routing table
1357 Destination Gateway Genmask Flags MSS Window irtt Iface
1358 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0
1359 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1
1360 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0
1361 127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo
1363 This routing configuration will likely still update the
1364 receive/transmit times in the driver (needed by the ARP monitor), but
1365 may bypass the bonding driver (because outgoing traffic to, in this
1366 case, another host on network 10 would use eth0 or eth1 before bond0).
1368 The ARP monitor (and ARP itself) may become confused by this
1369 configuration, because ARP requests (generated by the ARP monitor)
1370 will be sent on one interface (bond0), but the corresponding reply
1371 will arrive on a different interface (eth0). This reply looks to ARP
1372 as an unsolicited ARP reply (because ARP matches replies on an
1373 interface basis), and is discarded. The MII monitor is not affected
1374 by the state of the routing table.
1376 The solution here is simply to insure that slaves do not have
1377 routes of their own, and if for some reason they must, those routes do
1378 not supersede routes of their master. This should generally be the
1379 case, but unusual configurations or errant manual or automatic static
1380 route additions may cause trouble.
1382 8.2 Ethernet Device Renaming
1383 ----------------------------
1385 On systems with network configuration scripts that do not
1386 associate physical devices directly with network interface names (so
1387 that the same physical device always has the same "ethX" name), it may
1388 be necessary to add some special logic to either /etc/modules.conf or
1389 /etc/modprobe.conf (depending upon which is installed on the system).
1391 For example, given a modules.conf containing the following:
1394 options bond0 mode=some-mode miimon=50
1400 If neither eth0 and eth1 are slaves to bond0, then when the
1401 bond0 interface comes up, the devices may end up reordered. This
1402 happens because bonding is loaded first, then its slave device's
1403 drivers are loaded next. Since no other drivers have been loaded,
1404 when the e1000 driver loads, it will receive eth0 and eth1 for its
1405 devices, but the bonding configuration tries to enslave eth2 and eth3
1406 (which may later be assigned to the tg3 devices).
1408 Adding the following:
1410 add above bonding e1000 tg3
1412 causes modprobe to load e1000 then tg3, in that order, when
1413 bonding is loaded. This command is fully documented in the
1414 modules.conf manual page.
1416 On systems utilizing modprobe.conf (or modprobe.conf.local),
1417 an equivalent problem can occur. In this case, the following can be
1418 added to modprobe.conf (or modprobe.conf.local, as appropriate), as
1419 follows (all on one line; it has been split here for clarity):
1421 install bonding /sbin/modprobe tg3; /sbin/modprobe e1000;
1422 /sbin/modprobe --ignore-install bonding
1424 This will, when loading the bonding module, rather than
1425 performing the normal action, instead execute the provided command.
1426 This command loads the device drivers in the order needed, then calls
1427 modprobe with --ignore-install to cause the normal action to then take
1428 place. Full documentation on this can be found in the modprobe.conf
1429 and modprobe manual pages.
1431 8.3. Painfully Slow Or No Failed Link Detection By Miimon
1432 ---------------------------------------------------------
1434 By default, bonding enables the use_carrier option, which
1435 instructs bonding to trust the driver to maintain carrier state.
1437 As discussed in the options section, above, some drivers do
1438 not support the netif_carrier_on/_off link state tracking system.
1439 With use_carrier enabled, bonding will always see these links as up,
1440 regardless of their actual state.
1442 Additionally, other drivers do support netif_carrier, but do
1443 not maintain it in real time, e.g., only polling the link state at
1444 some fixed interval. In this case, miimon will detect failures, but
1445 only after some long period of time has expired. If it appears that
1446 miimon is very slow in detecting link failures, try specifying
1447 use_carrier=0 to see if that improves the failure detection time. If
1448 it does, then it may be that the driver checks the carrier state at a
1449 fixed interval, but does not cache the MII register values (so the
1450 use_carrier=0 method of querying the registers directly works). If
1451 use_carrier=0 does not improve the failover, then the driver may cache
1452 the registers, or the problem may be elsewhere.
1454 Also, remember that miimon only checks for the device's
1455 carrier state. It has no way to determine the state of devices on or
1456 beyond other ports of a switch, or if a switch is refusing to pass
1457 traffic while still maintaining carrier on.
1462 If running SNMP agents, the bonding driver should be loaded
1463 before any network drivers participating in a bond. This requirement
1464 is due to the interface index (ipAdEntIfIndex) being associated to
1465 the first interface found with a given IP address. That is, there is
1466 only one ipAdEntIfIndex for each IP address. For example, if eth0 and
1467 eth1 are slaves of bond0 and the driver for eth0 is loaded before the
1468 bonding driver, the interface for the IP address will be associated
1469 with the eth0 interface. This configuration is shown below, the IP
1470 address 192.168.1.1 has an interface index of 2 which indexes to eth0
1471 in the ifDescr table (ifDescr.2).
1473 interfaces.ifTable.ifEntry.ifDescr.1 = lo
1474 interfaces.ifTable.ifEntry.ifDescr.2 = eth0
1475 interfaces.ifTable.ifEntry.ifDescr.3 = eth1
1476 interfaces.ifTable.ifEntry.ifDescr.4 = eth2
1477 interfaces.ifTable.ifEntry.ifDescr.5 = eth3
1478 interfaces.ifTable.ifEntry.ifDescr.6 = bond0
1479 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
1480 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1481 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
1482 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1484 This problem is avoided by loading the bonding driver before
1485 any network drivers participating in a bond. Below is an example of
1486 loading the bonding driver first, the IP address 192.168.1.1 is
1487 correctly associated with ifDescr.2.
1489 interfaces.ifTable.ifEntry.ifDescr.1 = lo
1490 interfaces.ifTable.ifEntry.ifDescr.2 = bond0
1491 interfaces.ifTable.ifEntry.ifDescr.3 = eth0
1492 interfaces.ifTable.ifEntry.ifDescr.4 = eth1
1493 interfaces.ifTable.ifEntry.ifDescr.5 = eth2
1494 interfaces.ifTable.ifEntry.ifDescr.6 = eth3
1495 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
1496 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1497 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
1498 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1500 While some distributions may not report the interface name in
1501 ifDescr, the association between the IP address and IfIndex remains
1502 and SNMP functions such as Interface_Scan_Next will report that
1505 10. Promiscuous mode
1506 ====================
1508 When running network monitoring tools, e.g., tcpdump, it is
1509 common to enable promiscuous mode on the device, so that all traffic
1510 is seen (instead of seeing only traffic destined for the local host).
1511 The bonding driver handles promiscuous mode changes to the bonding
1512 master device (e.g., bond0), and propagates the setting to the slave
1515 For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
1516 the promiscuous mode setting is propagated to all slaves.
1518 For the active-backup, balance-tlb and balance-alb modes, the
1519 promiscuous mode setting is propagated only to the active slave.
1521 For balance-tlb mode, the active slave is the slave currently
1522 receiving inbound traffic.
1524 For balance-alb mode, the active slave is the slave used as a
1525 "primary." This slave is used for mode-specific control traffic, for
1526 sending to peers that are unassigned or if the load is unbalanced.
1528 For the active-backup, balance-tlb and balance-alb modes, when
1529 the active slave changes (e.g., due to a link failure), the
1530 promiscuous setting will be propagated to the new active slave.
1532 11. Configuring Bonding for High Availability
1533 =============================================
1535 High Availability refers to configurations that provide
1536 maximum network availability by having redundant or backup devices,
1537 links or switches between the host and the rest of the world. The
1538 goal is to provide the maximum availability of network connectivity
1539 (i.e., the network always works), even though other configurations
1540 could provide higher throughput.
1542 11.1 High Availability in a Single Switch Topology
1543 --------------------------------------------------
1545 If two hosts (or a host and a single switch) are directly
1546 connected via multiple physical links, then there is no availability
1547 penalty to optimizing for maximum bandwidth. In this case, there is
1548 only one switch (or peer), so if it fails, there is no alternative
1549 access to fail over to. Additionally, the bonding load balance modes
1550 support link monitoring of their members, so if individual links fail,
1551 the load will be rebalanced across the remaining devices.
1553 See Section 13, "Configuring Bonding for Maximum Throughput"
1554 for information on configuring bonding with one peer device.
1556 11.2 High Availability in a Multiple Switch Topology
1557 ----------------------------------------------------
1559 With multiple switches, the configuration of bonding and the
1560 network changes dramatically. In multiple switch topologies, there is
1561 a trade off between network availability and usable bandwidth.
1563 Below is a sample network, configured to maximize the
1564 availability of the network:
1568 +-----+----+ +-----+----+
1569 | |port2 ISL port2| |
1570 | switch A +--------------------------+ switch B |
1572 +-----+----+ +-----++---+
1575 +-------------+ host1 +---------------+
1578 In this configuration, there is a link between the two
1579 switches (ISL, or inter switch link), and multiple ports connecting to
1580 the outside world ("port3" on each switch). There is no technical
1581 reason that this could not be extended to a third switch.
1583 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
1584 -------------------------------------------------------------
1586 In a topology such as the example above, the active-backup and
1587 broadcast modes are the only useful bonding modes when optimizing for
1588 availability; the other modes require all links to terminate on the
1589 same peer for them to behave rationally.
1591 active-backup: This is generally the preferred mode, particularly if
1592 the switches have an ISL and play together well. If the
1593 network configuration is such that one switch is specifically
1594 a backup switch (e.g., has lower capacity, higher cost, etc),
1595 then the primary option can be used to insure that the
1596 preferred link is always used when it is available.
1598 broadcast: This mode is really a special purpose mode, and is suitable
1599 only for very specific needs. For example, if the two
1600 switches are not connected (no ISL), and the networks beyond
1601 them are totally independent. In this case, if it is
1602 necessary for some specific one-way traffic to reach both
1603 independent networks, then the broadcast mode may be suitable.
1605 11.2.2 HA Link Monitoring Selection for Multiple Switch Topology
1606 ----------------------------------------------------------------
1608 The choice of link monitoring ultimately depends upon your
1609 switch. If the switch can reliably fail ports in response to other
1610 failures, then either the MII or ARP monitors should work. For
1611 example, in the above example, if the "port3" link fails at the remote
1612 end, the MII monitor has no direct means to detect this. The ARP
1613 monitor could be configured with a target at the remote end of port3,
1614 thus detecting that failure without switch support.
1616 In general, however, in a multiple switch topology, the ARP
1617 monitor can provide a higher level of reliability in detecting end to
1618 end connectivity failures (which may be caused by the failure of any
1619 individual component to pass traffic for any reason). Additionally,
1620 the ARP monitor should be configured with multiple targets (at least
1621 one for each switch in the network). This will insure that,
1622 regardless of which switch is active, the ARP monitor has a suitable
1626 12. Configuring Bonding for Maximum Throughput
1627 ==============================================
1629 12.1 Maximizing Throughput in a Single Switch Topology
1630 ------------------------------------------------------
1632 In a single switch configuration, the best method to maximize
1633 throughput depends upon the application and network environment. The
1634 various load balancing modes each have strengths and weaknesses in
1635 different environments, as detailed below.
1637 For this discussion, we will break down the topologies into
1638 two categories. Depending upon the destination of most traffic, we
1639 categorize them into either "gatewayed" or "local" configurations.
1641 In a gatewayed configuration, the "switch" is acting primarily
1642 as a router, and the majority of traffic passes through this router to
1643 other networks. An example would be the following:
1646 +----------+ +----------+
1647 | |eth0 port1| | to other networks
1648 | Host A +---------------------+ router +------------------->
1649 | +---------------------+ | Hosts B and C are out
1650 | |eth1 port2| | here somewhere
1651 +----------+ +----------+
1653 The router may be a dedicated router device, or another host
1654 acting as a gateway. For our discussion, the important point is that
1655 the majority of traffic from Host A will pass through the router to
1656 some other network before reaching its final destination.
1658 In a gatewayed network configuration, although Host A may
1659 communicate with many other systems, all of its traffic will be sent
1660 and received via one other peer on the local network, the router.
1662 Note that the case of two systems connected directly via
1663 multiple physical links is, for purposes of configuring bonding, the
1664 same as a gatewayed configuration. In that case, it happens that all
1665 traffic is destined for the "gateway" itself, not some other network
1668 In a local configuration, the "switch" is acting primarily as
1669 a switch, and the majority of traffic passes through this switch to
1670 reach other stations on the same network. An example would be the
1673 +----------+ +----------+ +--------+
1674 | |eth0 port1| +-------+ Host B |
1675 | Host A +------------+ switch |port3 +--------+
1676 | +------------+ | +--------+
1677 | |eth1 port2| +------------------+ Host C |
1678 +----------+ +----------+port4 +--------+
1681 Again, the switch may be a dedicated switch device, or another
1682 host acting as a gateway. For our discussion, the important point is
1683 that the majority of traffic from Host A is destined for other hosts
1684 on the same local network (Hosts B and C in the above example).
1686 In summary, in a gatewayed configuration, traffic to and from
1687 the bonded device will be to the same MAC level peer on the network
1688 (the gateway itself, i.e., the router), regardless of its final
1689 destination. In a local configuration, traffic flows directly to and
1690 from the final destinations, thus, each destination (Host B, Host C)
1691 will be addressed directly by their individual MAC addresses.
1693 This distinction between a gatewayed and a local network
1694 configuration is important because many of the load balancing modes
1695 available use the MAC addresses of the local network source and
1696 destination to make load balancing decisions. The behavior of each
1697 mode is described below.
1700 12.1.1 MT Bonding Mode Selection for Single Switch Topology
1701 -----------------------------------------------------------
1703 This configuration is the easiest to set up and to understand,
1704 although you will have to decide which bonding mode best suits your
1705 needs. The trade offs for each mode are detailed below:
1707 balance-rr: This mode is the only mode that will permit a single
1708 TCP/IP connection to stripe traffic across multiple
1709 interfaces. It is therefore the only mode that will allow a
1710 single TCP/IP stream to utilize more than one interface's
1711 worth of throughput. This comes at a cost, however: the
1712 striping often results in peer systems receiving packets out
1713 of order, causing TCP/IP's congestion control system to kick
1714 in, often by retransmitting segments.
1716 It is possible to adjust TCP/IP's congestion limits by
1717 altering the net.ipv4.tcp_reordering sysctl parameter. The
1718 usual default value is 3, and the maximum useful value is 127.
1719 For a four interface balance-rr bond, expect that a single
1720 TCP/IP stream will utilize no more than approximately 2.3
1721 interface's worth of throughput, even after adjusting
1724 Note that this out of order delivery occurs when both the
1725 sending and receiving systems are utilizing a multiple
1726 interface bond. Consider a configuration in which a
1727 balance-rr bond feeds into a single higher capacity network
1728 channel (e.g., multiple 100Mb/sec ethernets feeding a single
1729 gigabit ethernet via an etherchannel capable switch). In this
1730 configuration, traffic sent from the multiple 100Mb devices to
1731 a destination connected to the gigabit device will not see
1732 packets out of order. However, traffic sent from the gigabit
1733 device to the multiple 100Mb devices may or may not see
1734 traffic out of order, depending upon the balance policy of the
1735 switch. Many switches do not support any modes that stripe
1736 traffic (instead choosing a port based upon IP or MAC level
1737 addresses); for those devices, traffic flowing from the
1738 gigabit device to the many 100Mb devices will only utilize one
1741 If you are utilizing protocols other than TCP/IP, UDP for
1742 example, and your application can tolerate out of order
1743 delivery, then this mode can allow for single stream datagram
1744 performance that scales near linearly as interfaces are added
1747 This mode requires the switch to have the appropriate ports
1748 configured for "etherchannel" or "trunking."
1750 active-backup: There is not much advantage in this network topology to
1751 the active-backup mode, as the inactive backup devices are all
1752 connected to the same peer as the primary. In this case, a
1753 load balancing mode (with link monitoring) will provide the
1754 same level of network availability, but with increased
1755 available bandwidth. On the plus side, active-backup mode
1756 does not require any configuration of the switch, so it may
1757 have value if the hardware available does not support any of
1758 the load balance modes.
1760 balance-xor: This mode will limit traffic such that packets destined
1761 for specific peers will always be sent over the same
1762 interface. Since the destination is determined by the MAC
1763 addresses involved, this mode works best in a "local" network
1764 configuration (as described above), with destinations all on
1765 the same local network. This mode is likely to be suboptimal
1766 if all your traffic is passed through a single router (i.e., a
1767 "gatewayed" network configuration, as described above).
1769 As with balance-rr, the switch ports need to be configured for
1770 "etherchannel" or "trunking."
1772 broadcast: Like active-backup, there is not much advantage to this
1773 mode in this type of network topology.
1775 802.3ad: This mode can be a good choice for this type of network
1776 topology. The 802.3ad mode is an IEEE standard, so all peers
1777 that implement 802.3ad should interoperate well. The 802.3ad
1778 protocol includes automatic configuration of the aggregates,
1779 so minimal manual configuration of the switch is needed
1780 (typically only to designate that some set of devices is
1781 available for 802.3ad). The 802.3ad standard also mandates
1782 that frames be delivered in order (within certain limits), so
1783 in general single connections will not see misordering of
1784 packets. The 802.3ad mode does have some drawbacks: the
1785 standard mandates that all devices in the aggregate operate at
1786 the same speed and duplex. Also, as with all bonding load
1787 balance modes other than balance-rr, no single connection will
1788 be able to utilize more than a single interface's worth of
1791 Additionally, the linux bonding 802.3ad implementation
1792 distributes traffic by peer (using an XOR of MAC addresses),
1793 so in a "gatewayed" configuration, all outgoing traffic will
1794 generally use the same device. Incoming traffic may also end
1795 up on a single device, but that is dependent upon the
1796 balancing policy of the peer's 8023.ad implementation. In a
1797 "local" configuration, traffic will be distributed across the
1798 devices in the bond.
1800 Finally, the 802.3ad mode mandates the use of the MII monitor,
1801 therefore, the ARP monitor is not available in this mode.
1803 balance-tlb: The balance-tlb mode balances outgoing traffic by peer.
1804 Since the balancing is done according to MAC address, in a
1805 "gatewayed" configuration (as described above), this mode will
1806 send all traffic across a single device. However, in a
1807 "local" network configuration, this mode balances multiple
1808 local network peers across devices in a vaguely intelligent
1809 manner (not a simple XOR as in balance-xor or 802.3ad mode),
1810 so that mathematically unlucky MAC addresses (i.e., ones that
1811 XOR to the same value) will not all "bunch up" on a single
1814 Unlike 802.3ad, interfaces may be of differing speeds, and no
1815 special switch configuration is required. On the down side,
1816 in this mode all incoming traffic arrives over a single
1817 interface, this mode requires certain ethtool support in the
1818 network device driver of the slave interfaces, and the ARP
1819 monitor is not available.
1821 balance-alb: This mode is everything that balance-tlb is, and more.
1822 It has all of the features (and restrictions) of balance-tlb,
1823 and will also balance incoming traffic from local network
1824 peers (as described in the Bonding Module Options section,
1827 The only additional down side to this mode is that the network
1828 device driver must support changing the hardware address while
1831 12.1.2 MT Link Monitoring for Single Switch Topology
1832 ----------------------------------------------------
1834 The choice of link monitoring may largely depend upon which
1835 mode you choose to use. The more advanced load balancing modes do not
1836 support the use of the ARP monitor, and are thus restricted to using
1837 the MII monitor (which does not provide as high a level of end to end
1838 assurance as the ARP monitor).
1840 12.2 Maximum Throughput in a Multiple Switch Topology
1841 -----------------------------------------------------
1843 Multiple switches may be utilized to optimize for throughput
1844 when they are configured in parallel as part of an isolated network
1845 between two or more systems, for example:
1851 +--------+ | +---------+
1853 +------+---+ +-----+----+ +-----+----+
1854 | Switch A | | Switch B | | Switch C |
1855 +------+---+ +-----+----+ +-----+----+
1857 +--------+ | +---------+
1863 In this configuration, the switches are isolated from one
1864 another. One reason to employ a topology such as this is for an
1865 isolated network with many hosts (a cluster configured for high
1866 performance, for example), using multiple smaller switches can be more
1867 cost effective than a single larger switch, e.g., on a network with 24
1868 hosts, three 24 port switches can be significantly less expensive than
1869 a single 72 port switch.
1871 If access beyond the network is required, an individual host
1872 can be equipped with an additional network device connected to an
1873 external network; this host then additionally acts as a gateway.
1875 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
1876 -------------------------------------------------------------
1878 In actual practice, the bonding mode typically employed in
1879 configurations of this type is balance-rr. Historically, in this
1880 network configuration, the usual caveats about out of order packet
1881 delivery are mitigated by the use of network adapters that do not do
1882 any kind of packet coalescing (via the use of NAPI, or because the
1883 device itself does not generate interrupts until some number of
1884 packets has arrived). When employed in this fashion, the balance-rr
1885 mode allows individual connections between two hosts to effectively
1886 utilize greater than one interface's bandwidth.
1888 12.2.2 MT Link Monitoring for Multiple Switch Topology
1889 ------------------------------------------------------
1891 Again, in actual practice, the MII monitor is most often used
1892 in this configuration, as performance is given preference over
1893 availability. The ARP monitor will function in this topology, but its
1894 advantages over the MII monitor are mitigated by the volume of probes
1895 needed as the number of systems involved grows (remember that each
1896 host in the network is configured with bonding).
1898 13. Switch Behavior Issues
1899 ==========================
1901 13.1 Link Establishment and Failover Delays
1902 -------------------------------------------
1904 Some switches exhibit undesirable behavior with regard to the
1905 timing of link up and down reporting by the switch.
1907 First, when a link comes up, some switches may indicate that
1908 the link is up (carrier available), but not pass traffic over the
1909 interface for some period of time. This delay is typically due to
1910 some type of autonegotiation or routing protocol, but may also occur
1911 during switch initialization (e.g., during recovery after a switch
1912 failure). If you find this to be a problem, specify an appropriate
1913 value to the updelay bonding module option to delay the use of the
1914 relevant interface(s).
1916 Second, some switches may "bounce" the link state one or more
1917 times while a link is changing state. This occurs most commonly while
1918 the switch is initializing. Again, an appropriate updelay value may
1921 Note that when a bonding interface has no active links, the
1922 driver will immediately reuse the first link that goes up, even if the
1923 updelay parameter has been specified (the updelay is ignored in this
1924 case). If there are slave interfaces waiting for the updelay timeout
1925 to expire, the interface that first went into that state will be
1926 immediately reused. This reduces down time of the network if the
1927 value of updelay has been overestimated, and since this occurs only in
1928 cases with no connectivity, there is no additional penalty for
1929 ignoring the updelay.
1931 In addition to the concerns about switch timings, if your
1932 switches take a long time to go into backup mode, it may be desirable
1933 to not activate a backup interface immediately after a link goes down.
1934 Failover may be delayed via the downdelay bonding module option.
1936 13.2 Duplicated Incoming Packets
1937 --------------------------------
1939 It is not uncommon to observe a short burst of duplicated
1940 traffic when the bonding device is first used, or after it has been
1941 idle for some period of time. This is most easily observed by issuing
1942 a "ping" to some other host on the network, and noticing that the
1943 output from ping flags duplicates (typically one per slave).
1945 For example, on a bond in active-backup mode with five slaves
1946 all connected to one switch, the output may appear as follows:
1949 PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
1950 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
1951 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
1952 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
1953 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
1954 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
1955 64 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
1956 64 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
1957 64 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
1959 This is not due to an error in the bonding driver, rather, it
1960 is a side effect of how many switches update their MAC forwarding
1961 tables. Initially, the switch does not associate the MAC address in
1962 the packet with a particular switch port, and so it may send the
1963 traffic to all ports until its MAC forwarding table is updated. Since
1964 the interfaces attached to the bond may occupy multiple ports on a
1965 single switch, when the switch (temporarily) floods the traffic to all
1966 ports, the bond device receives multiple copies of the same packet
1967 (one per slave device).
1969 The duplicated packet behavior is switch dependent, some
1970 switches exhibit this, and some do not. On switches that display this
1971 behavior, it can be induced by clearing the MAC forwarding table (on
1972 most Cisco switches, the privileged command "clear mac address-table
1973 dynamic" will accomplish this).
1975 14. Hardware Specific Considerations
1976 ====================================
1978 This section contains additional information for configuring
1979 bonding on specific hardware platforms, or for interfacing bonding
1980 with particular switches or other devices.
1982 14.1 IBM BladeCenter
1983 --------------------
1985 This applies to the JS20 and similar systems.
1987 On the JS20 blades, the bonding driver supports only
1988 balance-rr, active-backup, balance-tlb and balance-alb modes. This is
1989 largely due to the network topology inside the BladeCenter, detailed
1992 JS20 network adapter information
1993 --------------------------------
1995 All JS20s come with two Broadcom Gigabit Ethernet ports
1996 integrated on the planar (that's "motherboard" in IBM-speak). In the
1997 BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
1998 I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
1999 An add-on Broadcom daughter card can be installed on a JS20 to provide
2000 two more Gigabit Ethernet ports. These ports, eth2 and eth3, are
2001 wired to I/O Modules 3 and 4, respectively.
2003 Each I/O Module may contain either a switch or a passthrough
2004 module (which allows ports to be directly connected to an external
2005 switch). Some bonding modes require a specific BladeCenter internal
2006 network topology in order to function; these are detailed below.
2008 Additional BladeCenter-specific networking information can be
2009 found in two IBM Redbooks (www.ibm.com/redbooks):
2011 "IBM eServer BladeCenter Networking Options"
2012 "IBM eServer BladeCenter Layer 2-7 Network Switching"
2014 BladeCenter networking configuration
2015 ------------------------------------
2017 Because a BladeCenter can be configured in a very large number
2018 of ways, this discussion will be confined to describing basic
2021 Normally, Ethernet Switch Modules (ESMs) are used in I/O
2022 modules 1 and 2. In this configuration, the eth0 and eth1 ports of a
2023 JS20 will be connected to different internal switches (in the
2024 respective I/O modules).
2026 A passthrough module (OPM or CPM, optical or copper,
2027 passthrough module) connects the I/O module directly to an external
2028 switch. By using PMs in I/O module #1 and #2, the eth0 and eth1
2029 interfaces of a JS20 can be redirected to the outside world and
2030 connected to a common external switch.
2032 Depending upon the mix of ESMs and PMs, the network will
2033 appear to bonding as either a single switch topology (all PMs) or as a
2034 multiple switch topology (one or more ESMs, zero or more PMs). It is
2035 also possible to connect ESMs together, resulting in a configuration
2036 much like the example in "High Availability in a Multiple Switch
2039 Requirements for specific modes
2040 -------------------------------
2042 The balance-rr mode requires the use of passthrough modules
2043 for devices in the bond, all connected to an common external switch.
2044 That switch must be configured for "etherchannel" or "trunking" on the
2045 appropriate ports, as is usual for balance-rr.
2047 The balance-alb and balance-tlb modes will function with
2048 either switch modules or passthrough modules (or a mix). The only
2049 specific requirement for these modes is that all network interfaces
2050 must be able to reach all destinations for traffic sent over the
2051 bonding device (i.e., the network must converge at some point outside
2054 The active-backup mode has no additional requirements.
2056 Link monitoring issues
2057 ----------------------
2059 When an Ethernet Switch Module is in place, only the ARP
2060 monitor will reliably detect link loss to an external switch. This is
2061 nothing unusual, but examination of the BladeCenter cabinet would
2062 suggest that the "external" network ports are the ethernet ports for
2063 the system, when it fact there is a switch between these "external"
2064 ports and the devices on the JS20 system itself. The MII monitor is
2065 only able to detect link failures between the ESM and the JS20 system.
2067 When a passthrough module is in place, the MII monitor does
2068 detect failures to the "external" port, which is then directly
2069 connected to the JS20 system.
2074 The Serial Over LAN (SoL) link is established over the primary
2075 ethernet (eth0) only, therefore, any loss of link to eth0 will result
2076 in losing your SoL connection. It will not fail over with other
2077 network traffic, as the SoL system is beyond the control of the
2080 It may be desirable to disable spanning tree on the switch
2081 (either the internal Ethernet Switch Module, or an external switch) to
2082 avoid fail-over delay issues when using bonding.
2085 15. Frequently Asked Questions
2086 ==============================
2090 Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2091 The new driver was designed to be SMP safe from the start.
2093 2. What type of cards will work with it?
2095 Any Ethernet type cards (you can even mix cards - a Intel
2096 EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes,
2097 devices need not be of the same speed.
2099 3. How many bonding devices can I have?
2103 4. How many slaves can a bonding device have?
2105 This is limited only by the number of network interfaces Linux
2106 supports and/or the number of network cards you can place in your
2109 5. What happens when a slave link dies?
2111 If link monitoring is enabled, then the failing device will be
2112 disabled. The active-backup mode will fail over to a backup link, and
2113 other modes will ignore the failed link. The link will continue to be
2114 monitored, and should it recover, it will rejoin the bond (in whatever
2115 manner is appropriate for the mode). See the sections on High
2116 Availability and the documentation for each mode for additional
2119 Link monitoring can be enabled via either the miimon or
2120 arp_interval parameters (described in the module parameters section,
2121 above). In general, miimon monitors the carrier state as sensed by
2122 the underlying network device, and the arp monitor (arp_interval)
2123 monitors connectivity to another host on the local network.
2125 If no link monitoring is configured, the bonding driver will
2126 be unable to detect link failures, and will assume that all links are
2127 always available. This will likely result in lost packets, and a
2128 resulting degradation of performance. The precise performance loss
2129 depends upon the bonding mode and network configuration.
2131 6. Can bonding be used for High Availability?
2133 Yes. See the section on High Availability for details.
2135 7. Which switches/systems does it work with?
2137 The full answer to this depends upon the desired mode.
2139 In the basic balance modes (balance-rr and balance-xor), it
2140 works with any system that supports etherchannel (also called
2141 trunking). Most managed switches currently available have such
2142 support, and many unmanaged switches as well.
2144 The advanced balance modes (balance-tlb and balance-alb) do
2145 not have special switch requirements, but do need device drivers that
2146 support specific features (described in the appropriate section under
2147 module parameters, above).
2149 In 802.3ad mode, it works with systems that support IEEE
2150 802.3ad Dynamic Link Aggregation. Most managed and many unmanaged
2151 switches currently available support 802.3ad.
2153 The active-backup mode should work with any Layer-II switch.
2155 8. Where does a bonding device get its MAC address from?
2157 If not explicitly configured (with ifconfig or ip link), the
2158 MAC address of the bonding device is taken from its first slave
2159 device. This MAC address is then passed to all following slaves and
2160 remains persistent (even if the first slave is removed) until the
2161 bonding device is brought down or reconfigured.
2163 If you wish to change the MAC address, you can set it with
2164 ifconfig or ip link:
2166 # ifconfig bond0 hw ether 00:11:22:33:44:55
2168 # ip link set bond0 address 66:77:88:99:aa:bb
2170 The MAC address can be also changed by bringing down/up the
2171 device and then changing its slaves (or their order):
2173 # ifconfig bond0 down ; modprobe -r bonding
2174 # ifconfig bond0 .... up
2175 # ifenslave bond0 eth...
2177 This method will automatically take the address from the next
2178 slave that is added.
2180 To restore your slaves' MAC addresses, you need to detach them
2181 from the bond (`ifenslave -d bond0 eth0'). The bonding driver will
2182 then restore the MAC addresses that the slaves had before they were
2185 16. Resources and Links
2186 =======================
2188 The latest version of the bonding driver can be found in the latest
2189 version of the linux kernel, found on http://kernel.org
2191 The latest version of this document can be found in either the latest
2192 kernel source (named Documentation/networking/bonding.txt), or on the
2193 bonding sourceforge site:
2195 http://www.sourceforge.net/projects/bonding
2197 Discussions regarding the bonding driver take place primarily on the
2198 bonding-devel mailing list, hosted at sourceforge.net. If you have
2199 questions or problems, post them to the list. The list address is:
2201 bonding-devel@lists.sourceforge.net
2203 The administrative interface (to subscribe or unsubscribe) can
2206 https://lists.sourceforge.net/lists/listinfo/bonding-devel
2208 Donald Becker's Ethernet Drivers and diag programs may be found at :
2209 - http://www.scyld.com/network/
2211 You will also find a lot of information regarding Ethernet, NWay, MII,
2212 etc. at www.scyld.com.
git.openpandora.org / pandora-kernel.git / blob